Research on Collaborative Filtering Recommendation Method Based on User Behavior Analysis

Abstract

In online resource search systems, users who lack clear query or viewing objectives often have a certain degree of blindness. Users hope to obtain recommended items from other users with similar behaviors, because the filtered items are more likely to be accepted by new users. Based on this, the paper constructs a knowledge graph using the ontology concept architecture of the project as the knowledge representation foundation. It conducts in-depth research on user behavior around user behavior ontology, project resource ontology, user relationship graph instances, user similarity, and rating matrix completion, in order to find the target user's similar neighbor set. Based on the user's collaborative filtering recommendation algorithm, the predicted rating of the recommended project is calculated and Top-n projects are recommended to the target user.  The experimental evaluation results indicate that the collaborative filtering recommendation method based on user behavior analysis proposed in this study has good recommendation quality in online resource recommendation systems.